หน้าหลัก

จาก wiki.surinsanghasociety
ไปยังการนำทาง ไปยังการค้นหา

To conclude the discussion, we need to anxiety that forcing the users to remain inside a fixed location during the training phase could reduce the naturalness and realism of the scenario. This was maintained by making certain that the user joints had been inside the field of view in the Kinect through the training course of action. Even so, as a consequence, most users tended to location themselves close to the geometric center from the region and barely moved from there. We're at present contemplating the possibility of a far more natural environment to enable scenes exactly where the user could be situated in some positions in which the robot could possibly not see a number of the joints of your user. On the other hand, we program to permit the robot to track the user by moving itself, so it could adapt towards the altering circumstances of your scene, for example the user standing closer to or additional in the robot, and so forth. six. Conclusions This paper presented a method to endow a social robot with all the capacity to understand interactively by keeping a organic conversation with its human teacher. The all-natural interaction is achieved applying a grammar-based ASR, whose aim should be to recognize diverse sentences and to extract their semantic which means. Using the semantics as labels from the notion becoming discovered, the robot is in a position to understand customers which are not robotic specialists. Our program has been tested inside the application of pose recognition, in which the robot learns the poses adopted by the teacher, listening her explanations. Our experiment consisted of 24 non-robotics specialists coaching the robot nine unique poses in three instruction workouts. We evaluated our technique by comparing four learning algorithms, attaining satisfactory leads to all of them for the three workout routines. A robot with interactive finding out capabilities can adapt quickly to unique conditions, because the user can train it ad hoc for that circumstance. Furthermore, since the robot is capable of establishing organic interactions, the teacher does not require any knowledge in robotics. Despite the promising benefits, our system nonetheless presents a significant limitation. The maximum quantity of poses it may find out is restricted by the amount of semantics coded into the ASR's grammar. Additionally, these grammars are pre-written in a text file by the robot programmer. However, we already started operating on an extension to our program, targeted at solving this limitation. This extension consists of combiningSensors 2013,a grammar-based ASR with statistical language models. Combined, the user will be able to add new semantics to the grammar that could be employed to label the discovered concept, at the same time. On top of that, our function leaves other paths open for exploration. Firstly, in the HRI point of view, this paper has focused on the HRI in the robot's point of view. It remains to study how customers perceive what the robot has discovered and how this truth alterations their relation and their expectations towards it. Much more, understanding what the user thinks concerning the finding out process could result in superior education scenarios that would finish in robots that understand much better in the users. Secondly, this work opens the door for constructing a continuous finding out framework, where the robot actively seeks for new examples and asks inquiries of its teacher in regards to the concepts being discovered.